IBS

Analysis of IBS-Smart/Trio-Smart vs Control on IBS-specific medications

Medications by Encounters

Sankey graph

Intervention (IBS/TRIO/Both)

download (.png) download interactive (.html)

Flow of Drug Classes Between Clinical Encounters in the Intervention Group. The Sankey diagram illustrates longitudinal transitions in medication class use across six successive clinical encounters among participants in the intervention group. Each vertical column represents one encounter (rx1 through rx6), and each colored node corresponds to a distinct drug class, including gastro‑motility agents, anti-infective agents, antidepressants, anxiolytics, antiallergics, proton pump inhibitors/H2 blockers (PPI/H2B), and an “other” category.
The width of each flow is proportional to the number of participants whose medication class changed or remained stable between encounters.
Early encounters (rx1–rx2) show substantial initiation of gastro‑motility agents and PPI/H2B therapies, while anti-infective and antiallergic agents appear less frequently. Across encounters, the “other” medication category remains a dominant pathway, indicating frequent transitions into heterogeneous or infrequently used drug classes.
By later encounters (rx5–rx6), gastro‑motility agents again form a major proportion of flows, suggesting recurrent or persistent prescribing patterns in this class.
The diagram highlights both persistence (e.g., continuing gastro‑motility therapy across multiple encounters) and variability in medication use, with numerous cross‑class transitions.

Control

download (.png) download interactive (.html)

Similar graph for the control group is shown above. These graphs reveal distinct patterns in prescribing behavior and medication stability between the intervention (IBS-Smart, Trio-Smart, Combined IBS-Smart & Trio-Smart) and control groups.
Overall Pattern of Medication Transitions
Intervention Group:
The flows show more concentrated transitions, especially through gastro‑motility agents, anti‑infective agents, and PPI/H2B classes.
The “other” medication category is consistently prominent but becomes especially dominant around rx3–rx5, indicating broad shifts into miscellaneous or less frequent drug classes.
Control Group:
The flows appear more diffuse, with a wider variety of medication classes represented (e.g., antidiarrheal, analgesic).
Medications are more heterogeneous across encounters, suggesting less clustering into specific therapeutic classes.

IBS only

download (.png) download interactive (.html)

Trio-Smart only

download (.png) download interactive (.html)

IBS & Trio-Smart

download (.png) download (.html)

Transitional Probabilities

Intervention (IBS/TRIO/Both)

Transitional probabilities of medication switching for IBS/Trio/Both

from		to
		analgesic	antiallergic	antidepressant	antidiarrheal	antiinfective	anxiolytic	gastromotility	other	ppi/h2b	Total
analgesic	Count	0	0	0	0	0	0	0	1	0	1
	Row pct	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	100.0%	0.0%
antiallergic	Count	0	0	0	1	1	0	3	17	1	23
	Row pct	0.0%	0.0%	0.0%	4.3%	4.3%	0.0%	13.0%	73.9%	4.3%
antidepressant	Count	0	0	1	0	1	1	0	5	2	10
	Row pct	0.0%	0.0%	10.0%	0.0%	10.0%	10.0%	0.0%	50.0%	20.0%
antidiarrheal	Count	0	0	0	0	0	0	0	1	0	1
	Row pct	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	100.0%	0.0%
antiinfective	Count	1	0	0	0	7	1	12	10	2	33
	Row pct	3.0%	0.0%	0.0%	0.0%	21.2%	3.0%	36.4%	30.3%	6.1%
anxiolytic	Count	0	0	0	0	1	0	1	5	0	7
	Row pct	0.0%	0.0%	0.0%	0.0%	14.3%	0.0%	14.3%	71.4%	0.0%
gastromotility	Count	0	6	2	0	4	0	21	30	9	72
	Row pct	0.0%	8.3%	2.8%	0.0%	5.6%	0.0%	29.2%	41.7%	12.5%
other	Count	4	9	6	0	14	5	14	189	9	250
	Row pct	1.6%	3.6%	2.4%	0.0%	5.6%	2.0%	5.6%	75.6%	3.6%
ppi/h2b	Count	0	6	0	0	1	1	1	13	1	23
	Row pct	0.0%	26.1%	0.0%	0.0%	4.3%	4.3%	4.3%	56.5%	4.3%
Total	Count	5	21	9	1	29	8	52	271	24	420

The transition matrix for the intervention group quantitatively supports the Sankey visualization. Consistent with the prominent ‘other’ nodes and thick connecting bands in the Sankey, ‘other’ was the most frequent destination class (271/420 transitions, ~64.5%) and showed high persistence (other→other: 189/250, 75.6%).
+ Gastro‑motility agents were also common and exhibited partial stability (gastro→gastro: 21/72, 29.2%) with frequent transitions to ‘other’ (gastro→other: 30/72, 41.7%) and to PPI/H2B (9/72, 12.5%).
+ Anti‑infective transitions often moved toward gastro‑motility (12/33, 36.4%) or ‘other’ (10/33, 30.3%), consistent with the cross‑class flows shown in the Sankey.
+ Differences between the table and figure largely reflect that the table pools transitions across all encounter intervals, whereas the Sankey displays encounter-specific flow patterns. The Sankey is “time-resolved”; the table is “time-collapsed”. If a transition occurs mostly early (e.g., rx1→rx2) but not later, the Sankey will show that timing. The table will average it into one number.

Control

Transitional probabilities of medication switching for Controls

from		to
		analgesic	antiallergic	antidepressant	antidiarrheal	antiinfective	anxiolytic	gastromotility	other	ppi/h2b	Total
analgesic	Count	1	2	0	0	0	0	2	10	1	16
	Row pct	6.2%	12.5%	0.0%	0.0%	0.0%	0.0%	12.5%	62.5%	6.2%
antiallergic	Count	1	1	2	0	0	1	4	23	6	38
	Row pct	2.6%	2.6%	5.3%	0.0%	0.0%	2.6%	10.5%	60.5%	15.8%
antidepressant	Count	1	1	0	0	0	1	2	13	2	20
	Row pct	5.0%	5.0%	0.0%	0.0%	0.0%	5.0%	10.0%	65.0%	10.0%
antidiarrheal	Count	0	0	0	0	1	2	1	5	0	9
	Row pct	0.0%	0.0%	0.0%	0.0%	11.1%	22.2%	11.1%	55.6%	0.0%
antiinfective	Count	0	0	1	1	8	0	7	14	3	34
	Row pct	0.0%	0.0%	2.9%	2.9%	23.5%	0.0%	20.6%	41.2%	8.8%
anxiolytic	Count	2	1	0	0	0	0	0	9	4	16
	Row pct	12.5%	6.2%	0.0%	0.0%	0.0%	0.0%	0.0%	56.2%	25.0%
gastromotility	Count	0	5	3	2	4	2	12	34	6	68
	Row pct	0.0%	7.4%	4.4%	2.9%	5.9%	2.9%	17.6%	50.0%	8.8%
other	Count	14	18	8	3	20	9	18	232	22	344
	Row pct	4.1%	5.2%	2.3%	0.9%	5.8%	2.6%	5.2%	67.4%	6.4%
ppi/h2b	Count	1	8	4	1	3	2	4	27	5	55
	Row pct	1.8%	14.5%	7.3%	1.8%	5.5%	3.6%	7.3%	49.1%	9.1%
Total	Count	20	36	18	7	36	17	50	367	49	600

Medication transitions differed between groups. In the intervention group, movements into the “other” class were most frequent (64.5% of all transitions), with strong persistence within this category (other→other: 75.6%) and notable stability in gastro‑motility agents (gastro→gastro: 29.2%).

In contrast, the control group showed more heterogeneous switching patterns, including greater involvement of PPI/H2B and analgesic classes and lower stability within both “other” (67.4%) and gastro‑motility (17.6%).

Controls also exhibited more frequent transitions from and into PPI/H2B, whereas in the intervention group, anti‑infective agents more often shifted toward gastro‑motility (36.4% vs. 20.6%).

Overall, the transition matrices support the Sankey visualizations: the intervention group demonstrated more consistent medication trajectories, while the control group showed broader and less stable class‑to‑class movement.

In summary, compared with the intervention transition table, the controls show more PPI/H2B involvement, greater switching from gastro‑motility into “other,” and lower persistence within “other” and gastro‑motility. The intervention table shows greater continuity (especially other → other and gastro → gastro) and stronger anti‑infective → gastro‑motility transitions.

ibs-smart ONLY

Transitional probabilities of medication switching for IBS Only Group

from		to
		analgesic	antiallergic	antidepressant	antidiarrheal	antiinfective	anxiolytic	gastromotility	other	ppi/h2b	Total
analgesic	Count	0	0	0	0	0	0	0	1	0	1
	Row pct	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	100.0%	0.0%
antiallergic	Count	0	0	0	0	0	0	2	8	0	10
	Row pct	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	20.0%	80.0%	0.0%
antidepressant	Count	0	0	1	0	1	1	0	3	1	7
	Row pct	0.0%	0.0%	14.3%	0.0%	14.3%	14.3%	0.0%	42.9%	14.3%
antidiarrheal	Count	0	0	0	0	0	0	0	1	0	1
	Row pct	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	100.0%	0.0%
antiinfective	Count	0	0	0	0	5	1	7	4	0	17
	Row pct	0.0%	0.0%	0.0%	0.0%	29.4%	5.9%	41.2%	23.5%	0.0%
anxiolytic	Count	0	0	0	0	0	0	0	1	0	1
	Row pct	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	100.0%	0.0%
gastromotility	Count	0	2	2	0	3	0	17	19	8	51
	Row pct	0.0%	3.9%	3.9%	0.0%	5.9%	0.0%	33.3%	37.3%	15.7%
other	Count	3	3	4	0	6	0	8	112	6	142
	Row pct	2.1%	2.1%	2.8%	0.0%	4.2%	0.0%	5.6%	78.9%	4.2%
ppi/h2b	Count	0	4	0	0	1	0	0	9	1	15
	Row pct	0.0%	26.7%	0.0%	0.0%	6.7%	0.0%	0.0%	60.0%	6.7%
Total	Count	3	9	7	0	16	2	34	158	16	245

trio-smart ONLY

Transitional probabilities of medication switching for Trio-Smart Only Group

from		to
		analgesic	antiallergic	antidepressant	antiinfective	anxiolytic	gastromotility	other	ppi/h2b	Total
antiallergic	Count	0	0	0	1	0	1	6	1	9
	Row pct	0.0%	0.0%	0.0%	11.1%	0.0%	11.1%	66.7%	11.1%
antidepressant	Count	0	0	0	0	0	0	2	1	3
	Row pct	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	66.7%	33.3%
antiinfective	Count	1	0	0	0	0	1	4	1	7
	Row pct	14.3%	0.0%	0.0%	0.0%	0.0%	14.3%	57.1%	14.3%
anxiolytic	Count	0	0	0	1	0	1	4	0	6
	Row pct	0.0%	0.0%	0.0%	16.7%	0.0%	16.7%	66.7%	0.0%
gastromotility	Count	0	2	0	1	0	2	5	0	10
	Row pct	0.0%	20.0%	0.0%	10.0%	0.0%	20.0%	50.0%	0.0%
other	Count	1	4	2	4	5	2	58	3	79
	Row pct	1.3%	5.1%	2.5%	5.1%	6.3%	2.5%	73.4%	3.8%
ppi/h2b	Count	0	1	0	0	1	1	3	0	6
	Row pct	0.0%	16.7%	0.0%	0.0%	16.7%	16.7%	50.0%	0.0%
Total	Count	2	7	2	7	6	8	82	6	120

ibs-smart AND trio-smart

Transitional probabilities of medication switching for IBS & Trio-Smart Combined Group

from		to
		antiallergic	antidiarrheal	antiinfective	gastromotility	other	ppi/h2b	Total
antiallergic	Count	0	1	0	0	3	0	4
	Row pct	0.0%	25.0%	0.0%	0.0%	75.0%	0.0%
antiinfective	Count	0	0	2	4	2	1	9
	Row pct	0.0%	0.0%	22.2%	44.4%	22.2%	11.1%
gastromotility	Count	2	0	0	2	6	1	11
	Row pct	18.2%	0.0%	0.0%	18.2%	54.5%	9.1%
other	Count	2	0	4	4	19	0	29
	Row pct	6.9%	0.0%	13.8%	13.8%	65.5%	0.0%
ppi/h2b	Count	1	0	0	0	1	0	2
	Row pct	50.0%	0.0%	0.0%	0.0%	50.0%	0.0%
Total	Count	5	1	6	10	31	2	55

Propensity Score Matching

IBS-Smart

Summary

dt_test_prep <- dt_test_prep %>% mutate(ibssmart = case_when(is.na(ibs_smart)~0, TRUE~ibs_smart))                
mod_matchibs <- matchit(ibssmart ~ age+ history_before_2019+MCAS_first_visit+IBS_D_first_visit+refer_y, 
                     method = "nearest", 
                     data = dt_test_prep)
summary(mod_matchibs)

## 
## Call:
## matchit(formula = ibssmart ~ age + history_before_2019 + MCAS_first_visit + 
##     IBS_D_first_visit + refer_y, data = dt_test_prep, method = "nearest")
## 
## Summary of Balance for All Data:
##                     Means Treated Means Control Std. Mean Diff. Var. Ratio
## distance                   0.3529        0.2386          0.8266     0.8814
## age                       48.5254       52.6250         -0.2626     0.7040
## history_before_2019        0.3390        0.6438         -0.6438          .
## MCAS_first_visit           0.0000        0.0688         -0.3179          .
## IBS_D_first_visit          0.0339        0.0625         -0.1580          .
## refer_y                    0.7288        0.8500         -0.2726          .
##                     eCDF Mean eCDF Max
## distance               0.2111   0.3940
## age                    0.0693   0.1787
## history_before_2019    0.3048   0.3048
## MCAS_first_visit       0.0688   0.0688
## IBS_D_first_visit      0.0286   0.0286
## refer_y                0.1212   0.1212
## 
## Summary of Balance for Matched Data:
##                     Means Treated Means Control Std. Mean Diff. Var. Ratio
## distance                   0.3529        0.3411          0.0852     1.1474
## age                       48.5254       50.8475         -0.1488     0.8510
## history_before_2019        0.3390        0.3390          0.0000          .
## MCAS_first_visit           0.0000        0.0000          0.0000          .
## IBS_D_first_visit          0.0339        0.0508         -0.0937          .
## refer_y                    0.7288        0.7458         -0.0381          .
##                     eCDF Mean eCDF Max Std. Pair Dist.
## distance               0.0164   0.1186          0.0886
## age                    0.0395   0.0847          0.6548
## history_before_2019    0.0000   0.0000          0.1356
## MCAS_first_visit       0.0000   0.0000          0.0000
## IBS_D_first_visit      0.0169   0.0169          0.4683
## refer_y                0.0169   0.0169          0.3431
## 
## Sample Sizes:
##           Control Treated
## All           160      59
## Matched        59      59
## Unmatched     101       0
## Discarded       0       0

eQQ Plot

plot(mod_matchibs)

Covariate balance

love.plot(mod_matchibs)   

Figure presents a love plot summarizing covariate balance before (unadjusted; red) and after (adjusted; teal) the propensity-score–based adjustment for the IBS-Smart group and matched controls.

• The x‑axis shows the (standardized) mean difference between comparison groups for each covariate, with the vertical line at 0 indicating perfect balance.
  
• In the unadjusted sample, several baseline characteristics exhibited meaningful imbalance, with standardized mean differences generally in the range of approximately −0.1 to −0.35 for variables such as age, history_before_2019, MCAS_first_visit, IBS_D_first_visit, and refer_y_. In addition, the propensity score distance measure showed substantial separation between groups (large positive mean difference), consistent with strong baseline differences in treatment assignment prior to adjustment.
  
• After adjustment, the covariate mean differences moved markedly toward zero across all measured characteristics, indicating improved balance between groups. Most covariates cluster tightly around 0, suggesting the adjustment procedure was effective in reducing systematic baseline differences; the largest residual differences appear modest relative to the unadjusted sample.

TRIO-Smart

Summary

dt_test_prep <- dt_test_prep %>% mutate(triosmart_dv = case_when(is.na(trio_smart_dv)~0, TRUE~trio_smart_dv))                
mod_matchts <- matchit(triosmart_dv ~ history_before_2019 +ibs_first_visit +SIBO_first_visit  ,
                                      method = "nearest", data = dt_test_prep)
                 
summary(mod_matchts)

## 
## Call:
## matchit(formula = triosmart_dv ~ history_before_2019 + ibs_first_visit + 
##     SIBO_first_visit, data = dt_test_prep, method = "nearest")
## 
## Summary of Balance for All Data:
##                     Means Treated Means Control Std. Mean Diff. Var. Ratio
## distance                   0.2416        0.1853          0.7072      0.832
## history_before_2019        0.4186        0.5966         -0.3608          .
## ibs_first_visit            0.0465        0.1534         -0.5076          .
## SIBO_first_visit           0.0698        0.1989         -0.5067          .
##                     eCDF Mean eCDF Max
## distance               0.1184   0.2797
## history_before_2019    0.1780   0.1780
## ibs_first_visit        0.1069   0.1069
## SIBO_first_visit       0.1291   0.1291
## 
## Summary of Balance for Matched Data:
##                     Means Treated Means Control Std. Mean Diff. Var. Ratio
## distance                   0.2416        0.2416               0          1
## history_before_2019        0.4186        0.4186               0          .
## ibs_first_visit            0.0465        0.0465               0          .
## SIBO_first_visit           0.0698        0.0698               0          .
##                     eCDF Mean eCDF Max Std. Pair Dist.
## distance                    0        0               0
## history_before_2019         0        0               0
## ibs_first_visit             0        0               0
## SIBO_first_visit            0        0               0
## 
## Sample Sizes:
##           Control Treated
## All           176      43
## Matched        43      43
## Unmatched     133       0
## Discarded       0       0

eQQ Plot

plot(mod_matchts)

Covariate balance

love.plot(mod_matchts)   

Figure displays a covariate balance plot (love plot) comparing standardized mean differences between study groups before (unadjusted; red) and after (adjusted; teal) application of the propensity-score–based adjustment. The variables assessed include history_before_2019, ibs_first_visit, SIBO_first_visit, and the propensity score distance metric.

• In the unadjusted sample, several covariates show notable imbalance, with standardized mean differences ranging roughly from –0.20 to –0.35 for clinical history variables, and a large positive imbalance for the distance metric, indicating substantial baseline separation between groups prior to adjustment.
  
• After adjustment, all covariates shift sharply toward zero, with adjusted mean differences clustering tightly at or near 0. This pattern indicates that the adjustment procedure successfully improved covariate balance across all measured characteristics, substantially reducing baseline confounding.